Fluid topology optimisation for heat transfer: design for additive manufacturing
File(s)
Author(s)
Pietropaoli, Marco
Type
Thesis
Abstract
In this work, a Topology Optimisation (TO) method is developed to design efficient coolant systems for gas turbines. The method consists in an idealised sedimentation process by which a scalar variable, impermeability, is iteratively updated across a fluid dynamic domain. The impermeability uniquely determines the design solution, that is the solidified structures built to increase heat transfer while reducing pressure losses. The mathematical model is explicitly derived through a continuous adjoint approach, adopted to constrain the optimisation search by the fluids equations of motion and energy.
As a proof of concept, first results are compared on 2D benchmarks against state of the art techniques, showing reduced computational time and better performances of the final designs. The rest on the work focuses on 3-dimensional designs. High variability of results is found when varying weights of the cost functions and the parametrisation of the impermeability. In this respect, the Controlled Sedimentation technique is introduced to avoid premature convergence into local minima. Precisely, the technique aims to improve the robustness of the code and to increase the exploration of the search space leading the algorithm to hit lower value of the cost function.
Finally, a squared straight 3-dimensional domain is optimised under real turbine working conditions. The optimised geometry is validated through a high fidelity simulation performed in StarCCM+ including Detached Eddy Simulation (DES). Thermal efficiency is quantified through the wall Nusselt number, showing the efficacy of the geometry up to 15/20\% higher when compared to experimental results on ribs roughened surfaces.
The presented TO method produces solutions with a high level of structural complexity. In this sense, the final designs are suitable to fully exploit the flexibility offered by Additive Manufacturing, overcoming the design constraints imposed by classical manufacturing techniques as stamping and molding.
As a proof of concept, first results are compared on 2D benchmarks against state of the art techniques, showing reduced computational time and better performances of the final designs. The rest on the work focuses on 3-dimensional designs. High variability of results is found when varying weights of the cost functions and the parametrisation of the impermeability. In this respect, the Controlled Sedimentation technique is introduced to avoid premature convergence into local minima. Precisely, the technique aims to improve the robustness of the code and to increase the exploration of the search space leading the algorithm to hit lower value of the cost function.
Finally, a squared straight 3-dimensional domain is optimised under real turbine working conditions. The optimised geometry is validated through a high fidelity simulation performed in StarCCM+ including Detached Eddy Simulation (DES). Thermal efficiency is quantified through the wall Nusselt number, showing the efficacy of the geometry up to 15/20\% higher when compared to experimental results on ribs roughened surfaces.
The presented TO method produces solutions with a high level of structural complexity. In this sense, the final designs are suitable to fully exploit the flexibility offered by Additive Manufacturing, overcoming the design constraints imposed by classical manufacturing techniques as stamping and molding.
Version
Open Access
Date Issued
2018-10-01
Date Awarded
2019-01-01
Advisor
Montomoli, Francesco
Publisher Department
Aeronautics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)